Could you build an art-related service with help from AI? Possibly, but “I can organize the research” and “I can prove who painted this” are very different offers. A model’s confident answer is not a professional qualification or an authentication certificate.
If you enjoy art and are considering a new service in midlife, start by understanding the evidence an attribution requires. Then choose a task you can perform, check, and explain honestly. This article outlines research support, not a shortcut to becoming an art authenticator.
Separate the questions a client may be asking
Who made the work? What is its ownership history? What condition is it in? What might it sell for? These questions overlap, but an answer to one does not settle the others. A plausible artist attribution does not establish legal ownership, and a sales listing is not proof of authenticity.
The National Gallery’s account of identifying a painting by Raphael describes evidence from art-historical expertise, provenance, style, technique, and scientific examination. Its technical-analysis explanation describes methods including infrared reflectography and X-rays. A normal photograph does not contain all the information those investigations examine.
For formal needs such as insurance, an estate, or a disputed attribution, ask the receiving organization which expertise and documentation it requires. Do not promise that it will accept an AI-generated report.
Understand what an AI result actually tested
Research systems can analyze visual features in images. A 2018 study presented at AAAI, for example, examined stroke characteristics in 300 digitized drawings, mainly from a small group of artists. That is a specific research task with a defined dataset. It is not evidence that any chatbot can authenticate any painting uploaded to it.
When evaluating an AI service, ask which artists and types of work it covers, how reference works were selected, and whether testing used works excluded from training. Ask what errors occurred and what happens when an image is poor or outside the system’s scope. Independent expert interpretation and access to the actual object may still be necessary.
A result labeled “high confidence” needs an explanation of what that score means. Does it concern visual similarity, membership in a tested category, or something else? Do not silently translate it into “this object is genuine.” Ask the provider to explain the claim and its limits in writing.
Offer a deliverable you can check
A carefully bounded research service might organize documents supplied by a client, transcribe readable labels, compile a bibliography, or prepare questions for a relevant specialist. Define which sources you will examine and what you will deliver before accepting work. Avoid offering conclusions beyond your experience.
The Getty Provenance Index provides records of historic art ownership transfers, including dealer stock books and sales catalogs. Such resources can help investigate a lead. A matching title or surname alone does not show that a record concerns your client’s object.
For every possible match, retain the source, page or record identifier, date, quoted identifying details, and reasons for uncertainty. Compare the dimensions, medium, description, and ownership sequence where the documents supply them. Mark missing information explicitly.
From these supplied records, create a table of stated dates, names, object descriptions, and source references. Preserve the original wording of uncertain attributions. Flag contradictions and unreadable text. Do not infer missing owners, identify the artist, estimate value, or describe the work as authenticated.
Check every row against the source. AI can merge two people with similar names or turn an uncertain attribution into a definite one. The useful service is a traceable research record, not a more persuasive story.
A fictional example: a promising catalog entry
Imagine a client brings a landscape attributed by family tradition to a named artist. AI finds an auction entry with a similar title. The recorded dimensions differ, and there is no photograph or ownership link connecting the two.
Your report can document the possible lead and explain what remains unresolved. It cannot present that entry as the painting’s proven sales history. A relevant specialist may decide which further records or examination would help. This example is fictional and does not report an actual authentication.
Set honest business boundaries
Describe your experience accurately and agree on the scope, fee, deadline, and use of AI. State whether you provide document organization, research assistance, or a specialist opinion you are qualified to give. These should not be interchangeable labels.
Obtain permission to use client images and documents with the tools involved. Keep ownership records, addresses, valuations, and private correspondence out of general prompts unless the client has authorized that specific handling and the service is appropriate. Disclose any financial interest in a sale or referral.
Do not guarantee an attribution, a sale price, or income from the service. A small paid research task is still work that needs quality checks and clear limits. Our article on running a one-person business with AI explains how to track checking time, client approval, and the full cost of a job.
If your immediate aim is researching an object you own, begin with using AI to research an antique. For a professional art service, the next step is to define the evidence you can responsibly provide and the expertise you must bring in.
Updated September 24, 2026. Prepared with AI assistance and checked against the linked sources. Fictional examples illustrate a workflow and are not reports of product testing. See our disclaimer.
Featured photograph: Brett Wharton / Unsplash. Museum photograph shown for illustration; no authentication or appraisal of the pictured artworks was performed by this site.